Operational HWRF – Plans for 2014 and beyond.

Vijay Tallapragada & HWRF Team

Environmental Modeling Center, NCEP/NOAA/NWS, NCWCP, College Park, MD 20740.

HFIP Annual Review Meeting, February 19, 2014

1

Priorities for Operational HWRF for 2014 hurricane seasonAddress known problems/issues identified during the season:a) Weak storms continued posing significant challengeb) Moisture initialization in the model less than optimumc) Cold start (very first cycle for a numbered storm) cases behave different (worse) than the warm

startd) Land interactions and cold temperature bias over lande) Too small inner most domain to contain large stormsf) Insufficient vertical resolution for satellite data assimilationg) Coarse resolution of ocean model, inadequate conditions for choice of ocean domain in the

Atlantic, 1-D coupling in the East PacificFocus areas for development, testing and evaluation1. Increase the vertical resolution of atmospheric model to 61 levels with higher model top of 2 hPa2. Upgrade HWRF physics suite to include RRTM-G, Modified Ferrier microphysics, NOAH LSM. 3. Upgrade the ocean model (POM) to 1/12o MPI POM with unified trans-Atlantic basin and 3D

ocean for Eastern Pacific basin. Upgrade the coupler to run on multiple processors.4. Further improvements to HWRF vortex initialization scheme and HWRF Data Assimilation System5. Additional operational forecast products from HWRF to include simulated brightness

temperatures for new satellite sensors, several new variables for downstream applications and 9-minute ATCF output Many bug fixes and enhancements for the vortex tracker.

6. Pre-implementation tests based on proposed Q4FY13 GFS upgrades2

Weak storms still pose a significant challenge for HWRF.

Land interactions also impacted a few intensity forecasts for H. Ingrid

Potential initialization issues for cold start

Problematic Storms for Intensity Forecasts

3

Intensity forecasts for Cold Start

Cold Start Warm Start

Invest 95E became Priscilla 16E on 2013101406

Cold start (bogus) intensified the storm significantly

Warm start cycles well behaved

Could we cycle the vortex from Invest 95E for first cycle of Priscilla?

2013101406 2013101412

Improved initialization for Cold Start Cycles

4

Initialization upgrades1. Match the initial maximum wind speed over the lands2. Fix the bug of calculating south-west corner of initialization domain3. Storm center is used in the procedure instead of using parent domain center4. Remove the vorticity discontinuity along the filter domain5. Avoid cold starts for the first cycle of named/numbered storm through cycling of vortex from Invest cases

Preliminary results (Humberto)

2013 initialization Proposed upgraded init.

Strong intensity bias reduced

5

Operational HWRF (43level, pt=50mb) new HWRF (61level, pt=2mb)

Increased model vertical levels (43 to 61) with higher model top (50mb to 2mb)

6

Extended nest domainsHurricane Sandy (2012102718 +18hr fcst)

d02: 20% extended10°X10° to 12°X12°

d03: 10% extended6.5°X6.5° to 7.1°X7.1°

7

Data assimilation upgrades (Trial 1)1. Apply regional hybrid GSI analysis for both D1 and ghost domains2. Assimilate the conventional data and TDR, dropsondes (including inner core from aircraft

recon), GSPRO, satellite derived wind, brightness temperature from IR instruments (HIRS, AIRS, IASI, GOES Sounder) and MW instruments (AMSU-A, MHS, ATMS)

3. Set satellite thinning box to 90 KM for IR instruments, and 45 KM for MW instruments4. Increase model vertical levels to 61 and model top from 50mb to 2mb in order to

assimilate more satellite data5. Change 3-hourly FGAT to hourly FGAT – provide more accurate first guess fields,

especially for fast moving and developing storms (withdrawn due to resource issue)

Track Error intensity Error

Sat DA

OPR

Results from HFIP Stream 2.0 Satellite DA Impact Tiger Team

8

Infrastructure/DA upgrades Dynamics/Physics upgrades Final

H14A H14B H14C T14CNest

motion(H140)

NOAH LSM

(H141)

Upgraded Ferrier (H142)

RRTMG (H143)

Ocean(H144)

H214

Description

1. Sat Da with more vertical levels2. Extended d2/d33. Upgraded vortex initialization4. GSI upgrade5. Invest cycling

1. NoSat DA

Sat DA only for D01

Same as H14C except no DA in d01, use GFS analysis

New nest motion and high-freq.products

NOAH LSM

Separatespecies, Frimeadvection with other upgrades

RadiationMPI-POM with new coupler

Combinationof Best Performing components

*need to dotest runs with new GFS in WCOSS

Person All All All Sam Young Weiguo Chanh Zhan/URI All

Cases

Whole 2011,2012and 2013 storms2008, 09, 10 TDR cases

As in H14A

As in H14A

Prioritycases

Priority cases

Priority casesPriority cases

Priority cases

Whole 2011,2012and 2013 storm

Due date Feb. 15 Feb. 15 Feb. 15 Feb. 15 Feb. 15 Feb. 15 Feb. 15 Feb. 15 March 31

Platform Jet Jet Jet Jet Jet Jet Jet Jet Jet/WCOSS

2014 HWRF pre-implementation test plan

9

2010-2013 stats• 2014 HWRF Configuration with infrastructure/DA upgrades: • Increased vertical resolution from 42 to 61 levels• Increased model top from 50 hPa to 2 hPa• Expanded 9/3 km nest domains (about 20% more)• Modified vortex initialization • Improved DA using inner core TDR and dropsonde data (when available), and clear

sky satellite radiance data• Two configurations:

• H14C: Assimilate conventional data and satellite data in the outer domain using TC relocated GDAS forecasts as first guess

• T14C: Use GFS Analysis for parent domain (no GSI for outer domain), use TC relocated GDAS forecasts as first guess for high-resolution DA (ghost) domains

• T14C is chosen as the candidate for physics testing (4 separate experiments)• Radiation (RRTM-G); LSM (NOAH-LSM); Advected Ferrier Microphysics and MPIPOM-TC

• Expected benefits: • About 10-15% improvement in Atlantic Track and Intensity forecasts• Neutral or positive impact on Eastern Pacific Track and Intensity forecasts

• 4-season test results to follow (comparing H14C/T14C to 2013 HWRF)

10

2010-2013 ATL basin

2010-2013 EP basin

2010-2013 EP basin

What it takes in operations to run 2014 HWRF

• Resource requirements:• 30% more compute nodes (increase from 146 to 192)• 15 minutes additional run-time (increase from 75 to about 90 minutes)• Delivery time will move from t+5.45 to t+6.00 for each synoptic cycle

• Process optimization:• Process and job unification using python based scripts• Single pre-processing job and single post-processing job• Reduced I/O by eliminating several intermediate files• In-built swath generators for rainfall and max. wind (model will directly

accumulate the fields during the integration)• Fully functional GRIB2 support• Additional output includes variables requested by SPC, 2-D grids for

Hurricane Wave Models and 9-minute ATCF style storm vitals

HFIP Multi-Model Regional Ensemble Prediction System

1. 20-member HWRF Ensembles (Stream 1.5):• Use FY2014 HWRF configuration and Additional stochastic perturbations to microphysics• Use control member’s vortex for all ensemble members (no independent cycling) to

increase the reliability and on-time delivery of forecasts

2. Joint COAMPS-TC/HWRF/GFDL Multi-Model High-Resolution Regional Ensemble System (Stream 2.0):• NRL scientists (Doyle and Reinecke) in collaboration with EMC (Tallapragada and Zhang)

designed a plan for the HFIP multi-model ensemble using both HWRF and COAMPS-TC at 3 km horizontal resolution in an ensemble mode, consisting of at least 10 members for each model.

• A set of cases (approximately 460 in the W. Atlantic) will be used to evaluate the performance of the 3-km COAMPS-TC and HWRF ensemble performed in a retrospective mode.

• The GFDL ensemble system may be included in the analysis as well.• The joint HWRF and COAMPS-TC ensemble system will be demonstrated in real time during

the 2014 season for the W. Atlantic (likely 1 Aug-30 Oct). The ensemble forecasts will be displayed and made available on the web through HFIP.

HWRF based Stream 1.5/Stream 2.0 Real-Time Parallels

• HWRF-HYCOM Parallel Experiments (Stream 1.5 or Stream 2)• Conduct retrospective and real-time experiments

• High-Resolution Physics experiments• Continue exploring alternate physics suite for HWRF. Run

real-time demo of HWRF with advanced physics options• HWRF-POM/HYCOM-WaveWatchIII 3-way coupled system

experiments (Stream 2.0)• Collaborative effort with URI/GFDL. Conduct real-time demo

experiments for selected storms• HWRF for Global Tropical Oceans (Stream 1.0)

• Use FY2014 HWRF configuration (if fits on Jet)• Continue providing real-time forecasts for all storms

requested by JTWC (including SH)

Advancements to Operational HWRF – Transition to NMM-B/NEMS Multi-Scale Modeling System

• NCEP/AOML Collaborative effort supported by OAR Sandy Supplemental High Impact Weather Prediction Project (HIWPP) and leveraged by NOAA’s HFIP support

• Take advantage of NMMB in NEMS infrastructure for developing next generation global-to-local-scale modeling system for tropical cyclone forecasting needs and for comprehensive solutions for landfalling storms

• Planned development, testing and evaluation leading to potential transition to operations in the next 3-5 years

Scientific advancements include:• Scale aware and feature aware

physics for high-resolution domains and for multi-scale interactions

• Advanced techniques for inner core data assimilation with use all available aircraft recon data including TDR, FL, SFMR, and satellite radiance data

• High-resolution ensembles for prediction of RI/RW

• Enhanced land-air-sea-wave-hydrology coupled system

1717

System Current (Q3FY13) Q3FY14 Q3FY15 Q3FY18

Atmosphere Triple nested WRF NMM, storm centric

Triple nested WRF NMM, storm centric

Triple nested WRF, storm centric

High-resolution Hurricane nests within the basin-scale/global model (NMM-B/NEMS)

27:9:3km horizontal, 42 Levels

27:9:3km horizontal, 61 Levels, higher model top and expanded nested domains

18:6:2km horizontal, 64L (could be 15/5/1.67 to get closer to GFS resolution

2km or higher resolution hurricane nests with 128 Levels, global model top, with 10 member ensembles for each storm

Ocean POM (3D ATL and 1 D EPAC) 1/6o

resolution 23 levels

POM (Combined Trans-Atlantic domain at 1/12o

resolution, 23 levels and 3D ocean for East Pacific)

HYCOM (1/12o resolution 32 levels)

Global HYCOM (1/12o resolution, 100 levels)

Waves None None Wave Watch III Wave Watch III

Data Assimilation

One-Way Hybrid EnKF-3DVAR with vortex init., inner core NOAA-P3 TDR DA

One-Way Hybrid with inner core aircraft recon data (TDR/FL) and clear sky satellite radiance DA

One-Way Hybrid with inner core recon data (TDR/FL); clear and inner core cloudy radiance DA

Two-way hybrid 3D/4D En-Varwith inner core aircraft and all sky satellite radiance DA

Hurricane Physics

Ferrier Microphysics with explicit convection in 3km domain

Advanced Microphysics with high-resolution convection scheme, NOAH LSM and RRTMG Radiation

Advanced Microphysics, and land-air-sea-wave interactions

Scale and feature aware physics coupled to wave, hydrology, surge and inundation models

Basins NATL, EPAC, CPAC NATL, EPAC, CPAC NATL, EPAC, CPAC All Tropical Ocean basins

Max. storms 5 5 5 All existing tropical storms including genesis forecasts out to 7 days

HWRF Evolution to 2018

Future Outlook 5 to 25 years??????

.

5-10 years 2020+

Resolution/ Infrastructure

Basin-Scale HWRF with multiple moveable nests (at cloud resolving resolutions) and high-resolution HWRF ensemblesDownstream applications (including landfall related storm surge, waves, flooding and inundation)

Global to Local Scale Modeling to capture multi-scale interactionsHigh-Resolution Ensembles for events of interest

Physics Observations based physics Incorporate effects of sea-spray, aerosols, waves, boundary layer rolls –explicit representation of inner core processes

Ensemble based physics approach

DA/ Vortex Initialization

Hybrid/EnKF with 4-D VAR Vortex initialization within the DA, focus on assimilation of all-weather radiances and aircraft data

Part of the data assimilation for global system

Ocean/Wave/Land

Fully coupled ocean-wave-land-atmosphere system

Products & Downstream applications

Meeting the next-generation needs of Hurricane Specialists at NHC and JTWC

19

GFDL 2014 Hurricane Model Upgrade• Increased horizontal resolution of inner nest from 1/12th to

1/18th degree with reduced damping of gravity waves in advection scheme

• Improved specification of surface exchange coefficients (ch, cd) and surface stress computation in surface physics

• Improved specification of surface roughness and wetness over land.

• Modified PBL with variable Critical Richardson Number.• Advection of individual micro-physics species.• Improved targeting of initial storm maximum wind and storm

structure in initialization. (Reduces negative intensity bias in vortex specification)

• Remove of vortex specification for storms of 40 knots and less• Upgrade ocean model to 1/12th degree MPI POM with unified

trans-Atlantic basin and 3D ocean for Eastern Pacific basin• Remove global_chgres in analysis step (direct interpolation

from hybrid to sigma coordinates) 20

New Cd and Ch formulation

New ChNew CdCurrent HWRF and GFDL Cd

Current HWRF Ch

21

Comparison of New cd and ch with Recent Referenced Studies

Cd Ch

22

New GFDL Model Significantly Improved Intensity Skill at all Time Levels

23

24

Reduced Intensity Bias with New GFDL Model

25

Reduced Over-Intensification Tendency for Weaker Storm Intensity with New Model

Hurricane IsaacAugust 22nd, 12z, 2012

Current GFDLNew GFDL

Current GFDL

New GFDL

Current GFDLNew GFDL

Current GFDL

Hurricane KatiaAugust 29th, 18z, 2011

New GFDL

Current GFDLCurrent GFDL

Hurricane MariaSeptember 9th, 12z, 2011

New GFDL

Hurricane PhilippeSeptember 28th, 6z, 2011

Current GFDLNew GFDL

New GFDL

26

Improved Intensification for Developing Hurricanes with New GFDL Model

Hurricane DanielleAugust 23rd, 6z. 2010

Current GFDL

New GFDL

Hurricane EarlAugust 27th, 18z, 2010

Current GFDLNew GFDL

Current GFDLNew GFDL

New GFDL

Current GFDL

Current GFDLCurrent GFDL

Hurricane IgorSeptember 11th, 0z, 2010

New GFDL

Hurricane IkeSeptember 3rd, 12z, 2008

New GFDL

27

Modest Reduction in Track Errors

28

Largest Track Improvement for Developing Systems with New GFDL

Hurricane IreneAugust 21st, 12z, 2011

Tropical Storm LisaSeptember 22nd , 0z, 2010

Current GFDLNew GFDL

Current GFDL

New GFDL

Current GFDL

Hurricane OpheliaSeptember 22nd, 0z, 2011

Hurricane ErnestoAugust 6z, 0z, 19`1

New GFDL

Current GFDL

New GFDL

29

Current GFDL

Current GFDL

Current GFDLNew GFDL

New GFDL

New GFDL

New GFDL

Current HWRF

Current HWRF

Current HWRF

Current GFDL

Hurricane Dalila Hurricane Dalila

Hurricane Raymond Hurricane Raymond

Initial Time: July 1st, 12z, 2013 Initial Time: July 3rd , 0z, 2013

Initial Time: October 22nd , 0z, 2013 Initial Time: October 24th, 0z, 2013

Preliminary Eastern Pacific Results

30

Thanks for your attention

Questions?

Real-time and pre-implementation T&E HWRF products:

http://www.emc.ncep.noaa.gov/gc_wmb/vxt/index.html

Acknowledgements:

HWRF team at EMC

EMC and HFIP Management

Collaborations with NHC, DTC, HRD, GFDL, URI, CIRA and other HFIP partners

31